When an AI engine decides whether to cite a site, one of the quiet questions it weighs is: is there evidence behind this? A domain with one page — however polished — offers a single data point. A domain with fifty substantive pages that agree with each other offers a body of evidence.
What the measurements look like
In sites we've audited and tracked, the pattern is consistent: sites that get cited for their category tend to have a real corpus — service pages, guides, honest FAQs, often dozens to hundreds of URLs. One site we measured carried 324 URLs including 68 English articles and drew repeated citations; a technically flawless competitor with 216 URLs but zero articles drew none. Corpus wasn't the only difference, but a site with nothing beyond a homepage never even enters the comparison.
The honest version of the rule: there is no threshold number. Industries differ wildly. What's universal is the zero case — no content pages at all is a measurable, fixable gap. (We hold ourselves to this: this knowledge hub exists because our own site failed that check.)
Pages worth building, roughly in order
- One page per real service. Not one paragraph on a combined page — a page each, with what it costs, where it's offered, how long it takes. These answer buying questions directly.
- Guides that answer category questions. "How X works", "what Y costs in [city]", "how to choose Z". You're competing to be the source for questions your future customers ask before they know your name.
- An FAQ with real questions (see our FAQ guide for the craft).
- Proof pages: case write-ups with verifiable specifics, dated and honest about scope.
- Service × area pages — only where the distinction is real. "Aircon repair in Nonthaburi" as a page is legitimate if you genuinely serve Nonthaburi with distinct logistics, response times or teams worth stating. Twenty copies of one page with the city name swapped is template spam; engines detect the pattern, and it burns trust on everything else you publish.
What not to do
- Don't pad. Ten thin pages of rephrased filler are worse than three substantive ones — thin content dilutes the average quality signal of your whole domain.
- Don't let AI mass-generate your corpus unreviewed. Generated drafts are fine as raw material; unreviewed generated pages at scale converge on the same generic text as everyone else's, and add zero evidence.
- Don't stop. A corpus that grew steadily and then froze two years ago tells its own story. Sustainable pace beats a launch burst: a page or two a week, kept up, outperforms fifty pages in one month followed by silence.
The test for every page
Before publishing, ask: does this page contain at least one fact, answer or specific that exists nowhere else on my site? If yes, it adds evidence. If it exists to make the sitemap longer, it subtracts. Grow the site the way you'd want to be judged: more true things on record, nothing that's only there to be counted.